No Association between GPX1 Pro¹⁹⁸Leu and Breast Cancer Risk

Polymorphisms in the antioxidant selenoprotein glutathione peroxidase (GPX1) have been proposed to be associated with lung cancer (1) and breast cancer (2). In particular, the Leu allele of the Pro¹⁹⁸Leu polymorphism has been shown to have lower activity in MCF-7 cell lysates stimulated with sodium selenite (2). This polymorphism was strongly associated with risk of lung cancer (1); a significant association [odds ratio (OR), 1.9; 95% confidence interval (95% CI), 1.02-3.58] in a study of 79 breast cancer cases and 517 controls also has been observed (2). We assessed the association between GPX1 polymorphisms and breast cancer risk in the prospective Nurses' Health Study.

Detailed information regarding the design of this nested case-control study (cases, n = 1,323; controls, n = 1,910) has been published previously (3). In brief, 32,826 women who were free of diagnosed breast cancer were followed for incident disease from time of blood sample collection in 1989 and 1990 up to May 31, 2000. The two single nucleotide polymorphisms studied [−1040 G/A (rs3448) and Pro¹⁹⁸Leu (rs1050450)] were chosen as haplotype-tagging single nucleotide polymorphisms from the four single nucleotide polymorphisms in the National Cancer Institute's SNP500 database (http://snp500cancer.nci.nih.gov/home.cfm). Briefly, haplotype frequencies from 31 Caucasian samples were generated using the Phase program and haplotype-tagging single nucleotide polymorphisms were chosen using the BEST program (4). GPX1 genotyping analysis was done by the Taqman Allelic Discrimination method (Applied Biosystems, Foster City, CA) using primers and probe sequences available from the SNP500 database. Genotype-specific ORs and 95% CIs were calculated using unconditional logistic regression and were adjusted for age, age at menopause, postmenopausal hormone use, body mass index at age 18 years, weight gain since age 18 years, benign breast disease, and family history of breast cancer. ORs in relation to tumor stage and size were calculated in cases only, with <4 nodes involved and tumor size ≤2 cm being the reference categories. In these analyses, we compared carriers of the rare allele with noncarriers. Haplotype frequencies in pooled cases and controls were estimated using PROC HAPLOTYPE and linkage disequilibrium was calculated using PROC ALLELE in the SAS System version 8.0 (SAS Institute, Cary, NC). Regression substitution (5) and unconditional logistic regression (adjusted as above) were used for haplotype association testing, with the G-Pro haplotype as the reference category. Power was calculated using STPLAN (http://calculators.stat.ucla.edu/powercalc/).

Both polymorphisms were in Hardy-Weinberg equilibrium, and significant linkage disequilibrium existed between them (r = 0.4, D′ = 1.00; P < 0.001). There was no difference in genotype frequency between our control group and that of Hu et al. (P = 0.26, Fisher's exact test). No association was observed between either GPX1 genotype or the associated haplotypes and breast cancer risk (Table 1). No association was seen between GPX1 alleles and advanced disease [≥4 involved nodes (n = 81) compared with <4 nodes (n = 967) involved: −1040A OR, 1.37; 95% CI, 0.87-2.16; Leu¹⁹⁸ OR, 0.75; 95% CI, 0.48-1.19]. Additionally, rare GPX1 alleles were not significantly overrepresented among larger tumors [>2 cm (n = 245) compared with ≤2 cm (n = 964): −1040A OR, 1.09; 95% CI, 0.83-1.44; Leu¹⁹⁸ OR, 1.09; 95% CI, 0.83-1.45]. Lastly, no association was seen with any haplotype and tumor size or progression.

We did not observe an association between either the −1040A or the Leu¹⁹⁸ allele in the GPX1 gene and breast cancer risk among Caucasian women. We had >99% power to detect a significant relative risk of 1.9 (as reported by Hu et al.) for homozygous carriers of the Leu allele compared with women homozygous for the Pro allele. It is possible that the association between this allele and breast cancer risk shown by Hu et al. could actually be explained by loss of heterozygosity due to the use of tumor DNA as opposed to peripheral blood from patients in their study. However, no association between the Leu¹⁹⁸ allele and tumor progression or size was seen in the present study. Our study is >5 times larger than the study of Hu et al., suggesting that our result is less likely to be influenced by chance fluctuations in the case or control genotype frequency.